Abstract: Multiple Description Coding (MDC) is one of the promising methods for robust transmission over non-prioritized and unpredictable networks. Based on the Discrete Cosine Transform (DCT) and the Set Partition in Hierarchical Trees (SPIHT) compression method, this study proposes a new MDC framework. We make full use of the energy concentration of DCT and the similarity among the blocks composed of reordered DCT coefficients to apply the SPIHT algorithm to the transform-domain images composed of reordered DCT blocks. The purpose of using the reordered coefficients is to realize the energy redistribution. Redundancy is introduced by the full and partial encoding method which means the three descriptions, each using different bit rates to encode the information from three different orientations, i.e., vertical, horizontal and diagonal directions. For transmission we adopt three channels, each containing the hybrid information from three different directions. Experimental results demonstrate that present technique is effective and practical.